Related Experiment Video
Updated: Jun 28, 2026
Integrating Remote Sensing with Species Distribution Models; Mapping Tamarisk Invasions Using the Software for Assisted Habitat Modeling (SAHM)
Published on: October 11, 2016
IL-HS: a deep inception-LSTM architecture for enhanced lithological mapping using EnMAP hyperspectral remote sensing
Younes Khandouch1,2, Soufiane Hajaj3,4, Abderrazak El Harti5
1Laboratory of Metrology and Information Processing, Physics Department, Faculty of Sciences Agadir, Ibn Zohr University, B.P. 8106, 80000, Agadir, Morocco. khandouch.younes@edu.uiz.ac.ma.
A new deep learning model, Inception-LSTM Hyperspectral Mapper (IL-HS), significantly improves lithological mapping accuracy using hyperspectral satellite data. This advanced framework enhances mineral exploration and geoscientific understanding in complex terrains.
Area of Science:
- Geoscience
- Remote Sensing
- Artificial Intelligence
Background:
- Accurate lithological mapping is vital for geoscience and mineral exploration but challenging in complex, semi-arid regions.
- Hyperspectral satellite imagery offers rich spectral information but requires sophisticated processing for effective lithological classification.
Purpose of the Study:
- To introduce the Inception-LSTM Hyperspectral Mapper (IL-HS), a deep learning framework for enhanced lithological classification.
- To evaluate the performance of IL-HS using EnMAP hyperspectral data in a geologically complex area.
Main Methods:
- Developed a deep learning framework integrating InceptionV2 for spatial feature extraction and a bidirectional long short-term memory (Bi-LSTM) module for spectral information.
- Applied the IL-HS model to EnMAP hyperspectral data from the Kerdous inlier, Anti-Atlas, Morocco.
- Compared IL-HS performance against Support Vector Machines and 3D Convolutional Neural Networks.
Main Results:
- Achieved an overall accuracy of 98.05% across 26 lithological units, significantly outperforming existing models.
- Demonstrated perfect recall for copper and manganese formations and reliable distinction of spectrally similar units.
- Effectively mitigated spectral redundancy and mixing artifacts in heterogeneous and altered terrains.
Conclusions:
- The Inception-LSTM Hyperspectral Mapper (IL-HS) is a robust and scalable approach for hyperspectral lithological classification and mapping.
- IL-HS shows significant potential for geoscientific research, mineral resource assessment, and sustainable exploration.
- Deep learning models offer powerful solutions for complex remote sensing data analysis in geology.
